Turn a Movie Title Into a Narration Video With the narrator-ai-cli Skill

Movie-recap videos — the narrated "watch a whole film in minutes" format — are usually a multi-tool grind: script, clip data, voiceover, BGM, and a timeline, all assembled by hand. The narrator-ai-cli-skill repository packages that pipeline as an agent skill: you say "create a narration video for The Shawshank Redemption", and the agent runs the whole chain through the narrator-ai-cli tool and returns a download link.

Why This Skill Matters

This repo is the brain, not the hands. The SKILL.md it ships is a machine-readable instruction file that teaches agents how to operate the separate narrator-ai-cli command-line tool — the README's own analogy: the CLI is the hands, the skill is the brain; the CLI works alone in a terminal, while the skill never does. In the SKILL.md you get two workflow paths (Original Narration and Adapted Narration), three creation modes (Hot Drama, Original Mix, New Drama), resource-selection rules for the built-in library — roughly 100 movies, 146 BGM tracks, 63 dubbing voices, and 90-plus narration templates — plus all 18 documented API error codes with recommended actions, and cost estimation before a task is created.

Because it is plain markdown plus a references folder, the README's tested-platforms table covers OpenClaw, Windsurf, WorkBuddy, QClaw, Youdao Lobster, Yuanqi AI, Claude Code, and Cursor, with any markdown-reading agent listed as compatible.

Installation

Two installs are required, in this order. First the CLI (Python 3.10+, narrator-ai-cli v1.0.0+):

pip install "narrator-ai-cli @ git+https://github.com/NarratorAI-Studio/narrator-ai-cli.git"

Then configure the API key — and note this is a vendor-issued credential, not a self-serve signup: the README's only documented ways to get one are emailing the contact address or scanning the QR code on the repo page.

narrator-ai-cli config set app_key <your_app_key>

Finally, clone the skill into your agent's skills folder. For Claude Code (project-level), the documented path is:

mkdir -p /path/to/your/project/.skills
git clone https://github.com/NarratorAI-Studio/narrator-ai-cli-skill.git /path/to/your/project/.skills/narrator-ai-cli

The SKILL.md and the references/ directory are both required — clone the whole repo and do not flatten the files. (Cursor uses .cursor/rules/, OpenClaw uses ~/.openclaw/skills/; WorkBuddy and QClaw upload both parts through the skill-management UI instead.) To update later, run git pull inside the cloned directory.

Real Workflow: One Sentence to a Finished Recap

With the CLI installed, the key configured, and the skill in place, just describe the video:

用喜剧模板给《肖申克的救赎》做一支解说视频。

The README documents what happens next as a fixed pipeline: search the movie, select the template, choose BGM, pick a dubbing voice, generate the narration script, compose the video, and return a download link. The SKILL.md's agent rules require confirming with you before acting, so expect a quick back-and-forth on template and voice choices rather than a silent run — and per the fast-path and standard-path split, an original-narration job takes fewer steps than an adapted one.

The skill also handles standalone tasks outside full videos: voice cloning and text-to-speech, using the same configured key.

Tips

• Ask "show me what movies are available" first — the roughly 100-title library is the boundary of what a narration job can draw on.
• The SKILL.md documents cost estimation before task creation, so ask for the budget check when you anticipate a long video.
• Batch work is a documented use case: the README's example prompts include making five narration videos for five different action movies.
• The error-handling section maps all 18 API error codes to recommended actions — when a run fails, ask the agent to check the code against that table.

When Not to Use This

Everything runs through the vendor's API with an app key you must request, so if you need fully local processing or self-serve API access, this is not that tool. The workflow also presumes its template-and-resource library: narration in your own visual style, or movies outside the built-in catalog, is not what the documented paths describe.

See the leaderboard for more skills.